Rick R.

Key Skills

Software

No software listed

Top Subject Matter

AI safety/security evaluation for harmful or policy-violating content

Trust & safety enabling infrastructure for adversarial/harmful traffic detection and anomaly review

Legal Services & Contract Review

Top Data Types

Text

Document

Top Task Types

Segmentation

Classification

Red Teaming

Entity (NER) Classification

Freelancer Overview

Founder & AI Safety Researcher (AI safety evaluation and rubric-based output assessment for labeling workflows). Brings 4+ years of professional experience across legal operations, contract review, compliance, and structured analysis. Core strengths include PyRIT, garak, and JailbreakBench. Education includes Self-Managed Homelab, Nairobi, Infrastructure & Systems Administrator (2023). AI-training focus includes data types such as Text and labeling workflows including Evaluation, Rating, and Segmentation.

Labeling Experience

Founder & AI Safety Researcher - Independent

TextSegmentationClassification

Founded an AI safety and security evaluation practice focused on identifying jailbreak vectors, model misuse patterns, and adversarial prompt behavior using established testing frameworks. Designed and operated a self-hosted security monitoring stack to evaluate detection coverage against simulated attack scenarios. Applied rubric-based evaluations to assess model outputs for safety, policy compliance, and harmful-content risk while also researching defensive proxy and bot detection methods. • Evaluate LLM behavior for jailbreak and adversarial prompt risks using PyRIT, garak, and JailbreakBench • Build and run security monitoring tooling (Wazuh, Suricata, CrowdSec, Zeek, MISP) • Produce structured rationales aligned to safety and policy guidelines for consistent decisions • Manage and secure multi-server infrastructure (Proxmox, Docker, networking) supporting evaluation environments

2024 - Present

Founder & AI Safety Researcher (AI safety evaluation and rubric-based output assessment for labeling workflows)

Text

Led rubric-based AI safety evaluations by assessing model outputs for safety, policy compliance, and harmful-content risk, with an emphasis on consistent guideline application to ambiguous cases. Conducted adversarial testing by identifying jailbreak vectors, model misuse patterns, and harmful prompt behavior using established evaluation frameworks. Produced structured evaluative rationales suitable for downstream content review and AI training data labeling workflows. • Evaluated jailbreak and adversarial prompt behavior using PyRIT, garak, and JailbreakBench • Applied structured rubrics to rate or judge safety/policy adherence of outputs • Performed detection-coverage assessment via simulated attack scenarios • Developed defensive proxy/bot detection approaches to support policy enforcement decisions

2024 - Present

Infrastructure & Systems Administrator - Self-Managed Homelab

TextSegmentationEntity (NER) Classification

Architected and maintained a multi-server virtualization environment using Proxmox VE to host numerous isolated services with controlled access and resource boundaries. Managed DNS-layer content filtering and network policy enforcement using Pi-hole and AdGuard Home across a segmented home network. Configured authentication and API security practices for containerized services, including credential rotation and remediation of exposed secrets. • Maintain virtualization and container platforms (Proxmox, Docker) with strong segmentation and boundaries • Configure DNS filtering (Pi-hole, AdGuard Home) and network policy controls • Implement secure authentication and credential rotation for APIs and services • Set up observability pipelines (Prometheus, Loki, Grafana) for monitoring and anomaly detection

2023 - Present

Infrastructure & Systems Administrator (homelab support for defensive evaluation workflows)

Text

Supported AI safety and enforcement workflows by maintaining an infrastructure environment for monitoring, anomaly review, and defensive evaluation related to adversarial activity. Configured DNS-layer and network controls to detect or reduce harmful traffic patterns in a controlled research setting. Enabled cross-server observability that can inform review outcomes and detection criteria used for content or model safety assessments. • Managed monitoring/observability pipelines (Prometheus, Loki, Grafana) • Maintained network and DNS filtering controls (Pi-hole, AdGuard Home) • Configured security-focused container and authentication practices supporting evaluation workloads • Built infrastructure that facilitates simulated detection and anomaly review loops

2023 - Present

Education

Work History